Accuracy of continuous glucose monitoring in critically ill patients.

Objective: Hyperglycaemia requiring insulin infusions is common among critically ill patients. Attempts to tightly control glucose levels in the intensive care unit (ICU) have had mixed results partly due to hypoglycaemia. Continuous glucose monitoring (CGM) has been widely adopted among ambulatory persons with diabetes but tested only on small numbers of patients in the ICU. This study was undertaken to address the accuracy of CGM in a group of critically ill patients. Setting and outcomes: This observational study included critically ill ICU patients admitted between 2019 and 2023 to a tertiary referral, university-affiliated hospital with a single 19-bed ICU. Patients requiring insulin to control hyperglycaemia had a sensor (Dexcom G4 or G6) attached. Accuracy was assessed as bias and mean absolute relative difference (MARD). Results are presented as median (interquartile range). Results: 103 patients were assessed. The majority (97%) were mechanically ventilated and on vasopressors (92%), and a quarter required renal replacement therapy. Men comprised 68% of the cohort; median age was 63 years, and body mass index was 29.8. Sensors were applied for 79.5 (48-160) hours. The median bias was -0.30 mmol/L (-0.55, 0.05), and MARD was 11.23 (8.10-15.03). No clinical hypoglycaemia occurred with sensors in place. Conclusion: In 103 patients with multiorgan failure on insulin for hyperglycaemia, CGM had acceptable bias and MARD and no hypoglycaemia. Given the accuracy seen in this trial, CGM could be considered in future trials of glucose management and used as a safety measure to reduce hypoglycaemic events. Cinical trials registration: The study was not registered with the Australian and New Zealand Clinical Trials Registry).